EP0341444A2

EP0341444A2 - Rapid mutational analysis method

Info

Publication number: EP0341444A2
Application number: EP89106627A
Authority: EP
Inventors: Brian Dr. Seed; Andrew Peterson
Original assignee: General Hospital Corp
Current assignee: General Hospital Corp
Priority date: 1988-04-15
Filing date: 1989-04-13
Publication date: 1989-11-15
Anticipated expiration: 2009-04-13
Also published as: PT90265A; DE68919524T2; US5955264A; DE68919524D1; US5411861A; DK172589D0; EP0341444A3; ES2065933T3; DK172589A; ATE114821T1; EP0341444B1; JPH02203787A; PT90265B; US6579676B1; DK172110B1

Abstract

A rapid mutational analysis method for mapping protein epitopes is disclosed. This method has been used to identify the binding sites for 16 anti-CD2 and anti-CD4 monoclonal antibodies. The powerful, rapid, and simple method of the present invention allows isolation of a very large number of mutants, and is applicable to any surface protein for which a cDNA and monoclonal antibodies are available. The present method is especially useful in ligand binding site studies for the design of new ligands and drugs.

Description

Field of the Invention

This invention is in the field of molecular biology and immunology. It relates to a novel method of selecting for and analysing mutants. The present invention also relates to the use of this method to identify antigenic domains or epitopes of proteins or polypeptides.

Background

Resting human T cells bind sheep erythrocytes via a T cell specific 50kD cell surface protein called CD2 (Bach, J.F., et al., Transplantation 8:265-280 (1969); Howard, F.D., et al., J. Immunol. 126:2117-2122 (1981)). This phenomenon has long had practical utility, but, until recently, little known physiological significance. However, parallel studies of the interaction between T cells and sheep erythrocytes (Hunig, T., J. Exp. Med. 162:890-901 (9185); Hunig, T.R. J. Immunol. 136:2103-2108 (1986)), and T cells and their physiological targets (Shaw, S., et al., Nature 323:262-264 (1986)), have led to the identification of a specific molecular ligand for CD2 which is a widely distributed surface protein called, in the human case, LFA-3. CD2/LFA-3 interactions mediate cytolytic target conjugation (Shaw, S., et al., Nature 323:262-264 (9186)), thymocyte-epithelial adhesion (Vollger, et al., (1987)), and the mixed lymphocyte reaction (Martin, P.J., et al., J. Immunol. 131:180-185 (1983)). In addition, a broader role for the CD2 antigen has been suggested by the discovery that certain combinations of anti-CD2-monoclonal antibodies can directly activate mature T cells via an antigen-independent pathway.
At present, the most common method for mapping protein epitopes requires the synthesis of an array of short synthetic peptides spanning the protein sequence, and the use of these peptides in multiple binding assays (Geysen, H.M., et al., Science 235:1184-1190 (1987)). In order to identify specific residues important for antibody binding, variants of the peptide are synthesized with substitutions at each position. The synthetic peptide strategy has several limitations. If the antibody derives its affinity from interaction with disparate portions of the polypeptide backbone or with a novel conformation of the backbone, the peptide will be unable to mimic the entire protein in binding to the antibody. In order to identify individual residues contacted by the antibody, an extremely large number of peptide variants must be synthesized. The most exhaustive such study to date involved the assay of over 1500 individual peptides (Getzoff, E.D., et al., Science 235:1191-1196 (1987)).
Monoclonal antibodies have been used to select against viral envelope determinants (Yewdell, J.W., and Gerhard, W., Ann. Rev. Microbiol. 35:185-206 (1981)). Such selections are both less convenient and less sensitive than desired because the mutational alterations must be extracted from the viral genome, and mutations leading to viral inviability cannot be detected.

Summary of the Invention

The present invention relates to a rapid and simple method for mapping protein epitopes. The rapid mutational analysis technique of the present invention involves the selection of antigen cDNA mutations which lead to a loss of antigen-antibody reactivity. The method employs cDNA epitope-loss mutants and allows the sampling of a very large number of amino acid substitutions in the native molecule. The mutation frequency is high enough, with the mutagenesis method used, that rare variants can be efficiently isolated. The technique of the present invention is rapid and simple enough to allow a very large number of mutants to be isolated and can be applied to any surface protein for which a cDNA and monoclonal antibodies are available.
The ability to easily obtain a large number of epitope loss variants allows detailed mapping of the accessible surfaces of proteins, the identification of ligand binding sites, and the design of proteins that are less antigenic but have greater biological effect.
Moreover, using the methods of the present invention, it is possible to utilize suitably tailored ligands as negative selection reagents directly and, thus, extend the approach to allow identification of residues important for ligand or substrate/protein interaction which may occur in pockets inaccessible to antibodies. The identification of ligand binding sites will facilitate structural studies leading to the design of new ligands and drugs.
The method of the present invention has been used to define the regions through which the CD2 antigen binds to anti-CD2 monoclonal antibodies and to define the binding sites on the CD4 antigen for the human immunodeficiency virus (HIV).
CD2 cDNA mutations were selected which lead to loss of CD2-antibody reactivity. The pattern of amino acid substitutions in the mutants defines three distinct regions of the CD2 molecule: one epitopic region recognized by group I and II antibodies; a second epitopic region recognized by group III antibodies; and a third epitopic region recognized by group IV antibodies. Comparison of amino acid residues important for antibody binding and amino acid residues important for LFA-3 binding indicates that group I and II antibodies interact with one portion of the LFA-3 binding site; that group III antibodies interact with another portion of the LFA-3 binding site; and that group IV antibodies interact with still another portion, which is not involved in LFA-3 binding. In addition, the close correspondence between the effects of individual substitutions on group I antibody and LFA-3 binding suggests that group I antibodies mediate their effect on T cell activation by mimicking the effects of LFA-3 binding.

Brief Description of the Drawings

Figure 1 is a schematic representation of the mutant isolation vector and the procedure for mutant isolation.
Figure 1A is a representation of the vector piH3MCD2 used for isolation of mutants.
Figure 1B is a schematic representation of the strategy for isolation of mutants. In step (i) COS cells expressing mutant and wild-type molecules are treated with a monoclonal antibody recognizing the epitope whose loss is desired. Those cells which still express the epitope bind the antibody, which leads to subsequent complement fixation (step ii) and cell lysis. In step iii the remaining cells are treated with antibody recognizing a second epitope (iii) of CD2 and allowed to adhere to dishes coated with antisera recognizing mouse monoclonals. Only those cells which express the second epitope bind to the dishes (iv). Plasmid DNA is recovered from the adherent cells. The first step removes those molecules which still express the epitope whose loss is desired; the second step increases the efficiency of the procedure and ensures that null mutants are not obtained.
Figure 2 indicates the location of epitopes on the primary CD2 sequence.
Figure 2A presents the predicted sequence of the mature CD2 protein. The transmembrane region is underlined with a dark bar. The antibodies used are shown along the left margin. The symbols under the primary sequence indicate the sensitivity of each antibody to changes at that position. A "O" indicates that a mutant was obtained with a substitution at that position or that indirect immunofluorescence of mutants obtained with another antibody showed a sensitivity to substitution at that position. A "+" indicates that a substitution at that position was tested and was not found to affect antibody reactivity. The "=" symbol means that only proline at that position affected reactivity.
Figure 3 presents the mutant collection defining epitope regions of CD2. The sequence of short stretches of the CD2 polypeptide which include each of the two major epitopic regions are shown. The amino acid substitution variants acquired by mutant selection are shown underneath each wild-type sequence. Figure 3A presents the sequence of epitope region 1 of the CD2 polypeptide. Figure 3B presents the sequence of epitope region 2 of the CD2 polypeptide. The columns on the right indicate (from left to right) the antibody used for negative selection and the antibody(ies) used for positive selection. "All 16" means that all 16 of the monoclonals in Figure 2 were combined for the positive selection step. "7 others" means that seven antibodies other than 35.1 recognizing the first epitopic region were combined for the positive selection step. The variants directly under the CD2 sequence were obtained by selecting mutants from a pool of plasmids mutagenized throughout the portion of the cDNA encoding the extracellular domain of the protein. The variants under the bars indicate mutants acquired by selection using plasmid mutagenized only by oligonucleotides covering the span of bars.
Figure 4 presents the location of epitopes on the primary CD4 sequence. The predicted sequence of the CD4 protein is shown, along with the proposed HIV binding site (overlined), the transmembrane region (underlined) and the amino acid positions where mutant substitution occurs(*).

Description of the Invention

The method of the present invention for mapping epitopes of cell-surface antigens makes it possible to map epitopes of any surface protein for which a cDNA and monoclonal antibodies are available. The method is carried out through the use of cDNA epitope-loss mutants, allows the sampling of a very large number of amino acid substitutions in the native (naturally occurring) molecule and, thus, makes it possible to map the accessible surfaces of proteins, identify ligand binding sites and design proteins which are less antigenic than their natural counterparts, but also have a greater biological effect.
By "cell surface antigen" is meant a protein that is present on the cell surface; in general, a cell surface antigen is transported through the intracellular membrane system to the cell surface. Such antigens are usually anchored to the cell surface membrane through a carboxyl terminal domain containing hydrophobic amino acids that lie in the lipid bilayer of the membrane. As described below, the method of the present invention has been used to identify the binding sites of the human T cell receptor (CD2 antigen) and to identify the HIV binding site on the CD4 antigen. These are represented, respectively, in Figures 2 and 4.

General Scheme of CD2 Mutant Isolation

Mutant isolation is carried out as represented in Figure 1. The method represented in Figure 1 is discussed in terms of its use in isolating CD2 epitope loss mutants. However, it is to be understood that it has also been used to identify CD4 epitope loss mutants and is generally applicable to other cell surface antigens.
As represented in Figure 1, CD2 epitope loss mutants were isolated as follows: COS cells were transfected with a pool of mutagenized plasmids, cultured for 48 hours, harvested and treated sequentially with an anti-CD2 monoclonal antibody (i.e., with a monoclonal antibody recognizing the epitope whose loss is desired), rabbit anti-mouse immunoglobulin antibody and complement. This step is referred to as the negative selection step and is represented as step (i) in Figure 1B.
Because spontaneous deletion mutants arise frequently in COS cells (Calos, M.P., Proc. Natl. Acad. Sci., USA, 80:3015-3019 (1983); Razzaque, A. et al., Proc. Natl. Acad. of Sci. USA, 80:3010-3014 (1983), a positive selection step was carried out as follows: the cells spared by complement treatment were treated with antibody(ies) recognizing a distinct CD2 epitope(s) and allowed to adhere to dishes coated with goat anti-mouse immunoglobulin antibody. Wysocki, L.J. and Sato, V.L., Proc. Natl. Acad. Sci. USA, 75:2844-2848 (1978). Antibodies used for isolation of epitope-loss mutants are shown in Table 1.

Plasmid DNA recovered from the adherent cells (Hirt, B., J. Mol. Biol. 26:365-369 (1967)) is next transformed into E. coli, amplified, and reintroduced into COS cells for further rounds as appropriate. At the end of the selection process, DNA from individual bacterial colonies is transfected into COS cells which are then scored for antibody binding.

TABLE 1

Antibodies Used for Epitope-Loss Mutant Isolation
Antibody	Isotope
9.6	IgG_2a
7E10	IgG_2b
MT910	IgG₁
MT110	IgG₁
95-5-49
35.1	IgG_2a
T11/3PT2H9	IgG₁
T11/3T4-8B5	IgG_2a
9-2	IgM
Nu-Ter	IgG₁
CLB-T11/1	IgG₁
*39B21	IgG_2a
TS1/8.1.1	IgG₁
F92-3All	IgG₁
9.1	IgG_2b
OCH217	IgM

*Antibody 39B21 is a rat monoclonal and all others are mouse antibodies

In application of the method of the present invention to map epitopes of the CD2 antigen, the vector piH3M was used for isolation of mutants. A segment of cDNA encoding for the cell-surface antigen is inserted into the vector piH3M, as described in Applicant's co-pending United States patent application Serial No. 160,416, filed February 25, 1988, and incorporated herein by reference. piH3M contains a plasmid origin of replication and a suppressor tRNA gene allowing replication and selection in E. coli. In addition, it contains replication origins from bacteriophage M13 and SV40 virus (Figure 1a). These allow for production of a single-stranded version of the plasmid and for amplification of the plasmid following introduction into cells expressing the proteins necessary for SV40 replication, i.e., COS monkey cells. Expression of cell cDNA surface antigen in COS cells is directed by the immediate early region promoter from the human cytomegalovirus. RNA splicing and 3′ end processing signals are found downstream from the cDNA and are derived from SV40 virus.
In isolating and defining the regions through which CD2 cell surface epitopes bind to anti-CD2 monoclonal antibodies, cDNA mutations were selected by the methods described herein.
The initial mutant isolation takes advantage of the high mutation rate experienced by DNAs transfected into tissue culture cells (Calos, M.P. et al., supra; Rassaque, A. et al., supra; and Miller, J.H. et al., EMBO J. 3:3117-3121 (1984)). A population of plasmids, mutagenized by passage through COS cells was recovered in E. coli. The mutagenized pool was subjected to three subsequent rounds of selection using monoclonal antibody (Mab) 9.6 for the negative selection and Mab 35.1 for positive selection, and a single mutant was isolated (Figure 2).
The isolated mutant had two nucleotide substitutions, changing Lys48 to Asn and Asp186 to Glu. (Hereinafter, the various mutants will be referred to by a wild-type residue/mutant residue convention, so that, e.g., Lys48Asn indicates that the lysine at position 48 has been replaced with an aspargine.) Separation of the two changes by oligonucleotide mutagenesis showed that Lys48Asn was solely responsible for the loss of antibody 9.6 binding. However, further attempts using other antibodies to isolate additional mutants from this pool were unsuccessful.
The pattern of amino acid substitutions in the mutants defines three distinct regions of the CD2 molecule comprising many sequence variants: antibodies that participate in activation and block erythrocyte adhesion bind to a first region; antibodies that only block adhesion bind to a second region; and antibodies that participate in activation but do not block adhesion bind to a third region.
The method of the present invention has also been used to isolate and define the regions through which CD4 cell surface epitopes bind to anti-CD4 monoclonal antibodies. Using the present method, amino acid substitution variants of CD4 have been isolated. Mutations which affect binding of HIV are found in an epitope cluster which includes the Leu3a epitope.
In the following detailed description, reference will be made to various methodologies known to those of skill in the art of recombinant genetics. Publications and other materials setting forth such known methodologies to which reference is made are incorporated herein by reference.
Standard reference works setting forth the general principles of recombinant DNA technology include Darnell, J.E., et al., Molecular Cell Biology, Scientific American Books, Inc., publisher, New York, New York (1986); Lewin, B.M., Genes II, John Wiley & Sons, New York (1985); Old, R.W., et al., Principles of Gene Manipulation: An Introduction to Genetic Engineering, Second Edition, University of California Press, Berkeley, California (1981); and Maniatis, T., et al, Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Harbor, New York (1982).

CD2

Random Mutagenesis and Selection for CD2 Epitope Mutants

To increase the likelihood of mutant isolation, a general method was devised for the high efficiency, random mutagenesis of any DNA. See, e.g., Botstein, D. and D. Shortle, Science 229, 1193-1201 (1985). A homogeneous and nonspecific substitution of all possible basepairs encoding the CD2 extracelluar domain is created from a collection of 20 overlapping 33-unit (33-mer) oligonucleotides synthesized with a mixture of 95% of the wild-type base at each position and 1.7% of each of the other three bases. The oligonucleotides are overlapped because the residue attached to the matrix during synthesis cannot be conveniently adulterated and because the efficiency of incorporation of mutations falling at the very end of the oligonucleotide is not known.
In order to maximize the efficiency of base pair substitutions, the procedure of Kunkel was used to incorporate the degenerate oligonucleotides into expression vector piH3MCD2 (Figure 1a). Kunkel, T.A., Proc. Natl. Acad. Sci. USA 82:488-492 (1985). Based on the degree of degeneracy of the oligonucleotides and the efficiency of their incorporation, it was estimated that approximately 40% of the plasmids resulting from transformation into E. coli contained at least one base substitution. Twenty separate mutagenesis reactions were performed and transformed into E. coli, and a portion of each resulting culture pooled to form a mutant stock.
When an aliquot of the mutant stock was subjected to the selective regime using Mab 9.6 for negative selection and Mab 35.1 for positive selection, 10-15% of the recovered plasmids were found to bear the desired phenotype. After two rounds, the desired mutants comprised 50-75% of the plasmid population. The same mutant stock was used for all subsequent mutant selections.

Location of Mutations

One-hundred fourteen (114) primary mutants were isolated, resulting in a collection of 50 different amino acid sequence variants. The results of the mutant selections are summarized in Figures 2 and 3. All of the variation falls in three discrete regions. Region 1 is centered about Lys48 and contains 47 mutations for all of the group I antibodies (9.6, 7E10, MT11D and MT91D), all but one of the group II antibodies, and one group III antibody (Figure 3a). Region 2 is centered about Gly95. In this second epitope region, 27 mutants were obtained using five different antibodies for negative selection (Figure 3b). The sole stimulatory antibody contacts a wider region than any of the other antibodies, but no other clear distinction can be made; each antibody gives a unique pattern of mutation. Most of the antibodies recognizing region 2 have little effect on T cell activation. Region 3 is represented by a single mutation which causes loss of reactivity with both of the group IV antibodies (9.1 and OCH217).
Figure 3a in particular shows the mutant collection defining epitope region 1. CD2 residues 42-56 are shown above the amino-acid substitution encoded by each mutant. The first column on the right shows the antibody used for negative selection. The second column shows the positive selection antibody(s). "All 16" indicates that all 16 monoclonals in Table 1 were combined and used for the positive selection step. "7 others" means that the seven antibodies other than 35.1 recognizing the first epitopic region were combined for the positive selection step. Variants directly under the CD2 sequence were obtained by selecting mutants from a pool of plasmids mutagenized throughout the extracellular domain of the protein. The variants under the bars indicate mutants acquired by the method of this invention using plasmids mutagenized only by oligonucleotides covering the span of the bars. The mutant collection defining epitope region 2 is shown in Figure 3b. CD2 residues 86 through 100 are shown above the mutant substitutions. Other notations are as in Figure 3a.

Erythrocyte Rosetting

The ability of the mutant CD2 proteins to promote LFA-3 mediated adhesion of human erythrocytes to transfected COS cells was measured by a qualitative erythrocyte rosette assay. Three phenotypes were scored; wild-type, partial, and non-rosetting. Many of the mutations leading to changes in region 1 and 2 (reactive with groups I, II and III antibodies) dramatically reduced rosetting. To examine this further, a few mutants were created by specific oligonucleotide mutagenesis. Substitution of asparagine or alanine for lysine at each of positions 46, 47, and 48 of epitope region 1 demonstrated a striking correlation between the binding of the group I antibody Mab 9.6 and erythrocyte adhesion Lys46Asn/Ala showed a modest effect on both Mab 9.6 and erythrocyte binding; Lys 47Asn/Ala had no effect on either; and Lys48Asn/Ala completely abolished both interactions. Similarly, residue 51 could be shown to be important for both erythrocyte and Mab 9.6 binding. Residue 52 had only a weak effect on either interaction.
Subsequent experiments with other antibodies and other mutants showed that Lys48 plays a major role in the interaction of CD2 with group I antibodies and LFA-3 (Figures 2 and 4). For example, the mutant Lys48Gly is unreactive with all of the group I antibodies, and none of the molecules substituted at Lys48 has any detectable rosetting activity. The behavior of substitutions at Lys48 constitutes one of the strongest pieces of evidence that group I antibodies mimic the effect of LFA-3 binding in provoking T cell proliferation.
Only one antibody which participates in the activation of T cells recognizes determinants in the second epitopic region. Substitutions in this area characteristically diminish rosetting without completely eliminating it, in agreement with the notion that mutations selected with antibodies which promote activation are usually mutations that affect rosetting.

Effects of Individual Amino Acid Substitutions

Although some residues can directly determine antibody reactivity, it appears that others, of secondary importance for antibody affinity, can be identified because they frequently are altered in association with other changes. For example, a Lys46Asn substitution frequently is found in mutants which do not bind Mab 9.6, but by itself causes only partial loss of binding. A similar phenomenon may be present in the repeated isolation of position 51, 52 double mutants.
In principle, amino acid substitutions could lead to loss of antibody or LFA-3 binding by either elimination of a specific interaction or by causing a local denaturation. Some patterns of antibody or LFA-3 binding argue against the latter possibility. For example, all of the molecules substituted at Lys48 still bind Mab 35.1, which is sensitive to changes at Ile49. Similarly, both Mab 9.6 and LFA-3 cannot bind to the Gln51Leu variant, which nonetheless is recognized by antibodies 7E10 and 9-2. CD2 having a Gln51Arg substitution is unreactive with 7E10 and 9-2 but binds LFA-3 indistinguishably from wild-type. In the second epitopic region, the Tyr91Asp mutation causes loss of rosetting, but antibody NU-TER binding is not affected, even though many substitutions at position 92 eliminate NU-TER reactivity.
In one case, local denaturation may play a prominant role in loss of antibody binding. Proline residues are known to have limited conformational freedom which does not favor alpha helix formation (Chou, P.Y. and Fasman, G.D., Ann. Rev. Biochem. 47:251-276 (1978)). Gln51Pro variants are not recognized by any of the antibodies reacting with the first epitopic region, and Gln51Pro is frequently isolated by negative selection with region 1 antibodies if positive selection with all 16 antibodies is used. In the most extreme case, negative selection with Mab 35.1 leads exclusively to Gln51Pro when all 16 antibodies are combined for positive selection. To avoid repetitive isolation of Gln51Pro, many of the mutants in the first epitopic region (Figures 2 and 3) were isolated using Mab 35.1 as the sole positive selection antibody.
To isolate a 35.1 mutant other than Gln51Pro, only the antibodies which fail to bind to this variant were used for positive selection. After three cycles of enrichment, a single 35.1 Ile49Gln mutant was obtained which was altered in all three bases of the original codon. The unusual nature of this mutation suggests that the 35.1 antibody derives its affinity from multiple features of the CD2 conformation, so that substitution for a single feature only rarely leads to greatly lowered affinity. The Gln51Pro mutation may eliminate several of these interactions by gross alteration of the local secondary structure. Because the affinity of the 35.1 antibody is comparable to that of antibody 9.6 (Martin, P.J. et al., J. Immunol. 131:180-185 (1983)), it appears that the unusual mutational spectrum of this antibody arises from a qualitatively different mode of binding and not simply a stronger interaction. Another group II antibody, T11/3PT2H9, also gave Gln51Pro mutations exclusively when all 16 monoclonals were pooled for the positive selection step.

Group IV Antibody Epitope

Only one mutant was obtained for the group IV antibodies, a Tyr140Asn/Gln141His double substitution. However, group IV antibodies react only weakly with the CD2 molecule expressed on COS cells, a situation reminiscent of the weak reactivity of group IV antibodies with CD2 on unactivated T cells, (Meuer, S.C., et al., Cell 36:897-906 (1984)). Prior activation of T cells or incubation with a group I antibody is necessary to make the group IV antibody epitope available (Meuer, S.C., et al., supra). The rapid acquisition of group IV antibody reactivity suggests that it is caused by a conformational change in the molecule and not by de novo synthesis of a different species (Meuer, S.C., et al., supra; Yang, S.Y., et al., J. Immunol. 137:1097-1100 (1986)).
Each of the monoclonal antibodies in this study gave a contiguous linear pattern of mutational variation. All three epitopic regions are hydrophilic as would be expected for an exposed portion of the molecule available for antibody binding (Figure 2). Several alternatives may be put forth to explain why only a few restricted portions of the molecule give rise to multiple independent antibodies. For example, a portion of the first epitope is predicted to form an alpha helix with hydrophilic residues on one side of the helix and hydrophobic residues on the other (Chou, P.Y., and Fasman, G.D., Ann. Rev. Biochem. 47:251-276 (1978)). Such a helix is thought to form a particularly favored antigen for T cell recognition (De Lisi, C., and Berzofsky, J.A., Proc. Natl. Acad. Sci. USA 82:7048 (1985)), and recognition of this region by mouse helper T cells may focus the antibody response. In the region corresponding to epitope region 2 (Figure 3b), three potential N-linked glycosylation sites are found in the rat CD2 sequence (Williams, A.F., et al., J. Exp. Med. 165:368-380 (1987)) which are not present in the human sequence. This may serve to enhance reactivity with the human sequence by reducing the number of mouse suppressor T cells which might cross-react with the human sequence. Alternatively, the restricted spectrum of antibody binding sites may arise from the prior selection of antibodies for erythrocyte receptor reactivity.

CD4

The method of the present invention has been used to isolate amino acid substitution variants of CD4. Mutations which affect binding of HIV are found in an epitope cluster which includes the Leu3a epitope (Figure 4). The epitope mapping experiments show that most anti-CD4 antibodies, including those which are most effective at blocking CD4-gp120 interactions, recognize the amino-terminal domain of CD4. The locations of epitopes together with the similarity between the amino-terminal domain of CD4 and immunoglobulin V domains allows modelling of the surface of CD4 which interacts with gp120.
CD2 may mediate both cell-cell adhesion and antigen-independent activation reactions. The HIV envelope protein gp120 binds to CD4 with high affinity (Kd 10⁻⁹M), allowing entry of the virus into the host cell. Cell surface expression of CD4 is necessary for viral penetration and appears, in human cells, to be sufficient for susceptibility to infection. Interaction between CD4 and gp120 also mediates syncytium formation, between infected cells and uninfected cells bearing CD4, which is at least partly responsible for the cytopathic effects observed following viral infection in vitro.
The functionally-defined binding sites identified by the method of this invention, and in particular those directed to the CD2 molecule or the CD4 molecule, can be prepared in soluble form and may therefore be used in immunodiagnostic assay methods well known in the art, including radio-immunoassays, enzyme immunoassays and enzyme-linked immunosorbent assays. Moreover, cell-surface antigens isolated by the method of the present invention can be prepared in soluble form and administered alone or in combination with other cell surface antigens of this invention for the treatment of immune-related disorders in animals, including humans. Examples of such conditions are immune deficiency diseases, AIDS, asthma, rheumatoid arthritis, immunopathogenic renal injury, immune endocrinopathies, and tissue/organ transplant rejection.
Soluble forms of CD4, lacking the third disulfide bonded extracellular domain as well as the transmembrane and cytoplasmic regions, are able to bind gp120. Deen, K.C. et al., Nature, 331:82-84 (1988); Dalgleish, et al., The Lancet, November 7, 1987, pp. 1047-1049. This limits the HIV binding activity of CD4 to the two most amino terminal domains. The most amino terminal domain has a great deal of homology to immunoglobulin V region domains, suggesting that gp120 may interact with an immunoglobulin-like domain of CD4.
The binding site for HIV on CD4 has been indirectly localized by using monoclonal antibodies to interfere with CD4-gp120 interactions. Two epitopes of CD4 seem to be near the HIV binding site, one defined by the Leu3a and OKT4A antibodies and the other defined by the MT151 and VIT4 antibodies. The two epitopes are spatially distinct, antibodies recognizing one epitope do not interfere with binding of antibodies recognizing the other. Anti-idiotype antibodies raised against the Leu3a antibody also recognize a conserved portion of gp120 indicating that the Leu3a epitope may actually comprise part of the HIV binding site.
Thus, as a result of identification of the HIV binding site on CD4, it is possible to produce peptides which interfere with the ability of HIV to infect human cells by preventing the virus from interacting with the CD4 cell surface receptor. This can be done, for example, by producing peptides which bind to the HIV, thus preventing it from binding to the cell surface receptor. Particularly useful, for example, for this purpose is a peptide having the same, or essentially the same, amino acid sequence as the Leu3a epitope, or that overlined in Figure 4. Such a peptide can be synthesized, using known techniques, and introduced (e.g., by intravenous administration) into an individual in soluble form (e.g., as part of another protein, such as an immunoglobulin) in sufficient quantitites to bind with the HIV and interfere with its ability to infect cells.
When used for immunotherapy, the antigens of the present invention can be labeled or unlabeled with a therapeutic agent. Examples of such agents include drugs, radioisotopes, lectins, and cell toxins. Methods and compositions of this invention will be further exemplified by the following, non-limiting examples.

EXAMPLE I Oligonucleotide Mutagenesis

Degenerate oligonucleotides were synthesized on an Applied Biosystems DNA synthesizer using a mixture of phosphoramidites, 95% of the wild-type sequence and 1.7% of each of the other three phosphoramidites, at each position. The 600 nucleotides of CD2 sequence following position 63 as it appears in Seed, B. and Aruffo, A., Proc. Natl. Acad. Sci. USA, 84:3365-3369 (1987) were synthesized in a collection of twenty 33-mer oligonucleotides. The sequence also appears in Applicants' co-pending United States patent application Serial No. 160,416, filed February 25, 1988. Each oligonucleotide overlaps with the preceding oligonucleotide sequence by three bases. The 3′ most base is immutable (the synthesis proceeds 3′-5′ from a resin fixed phosphoramidite); however, mutants can be obtained that are altered in the 5′-most base determined by an oligonucleotide. This would imply that an overlap of 1 base is sufficient to make all positions mutable. The oligonucleotides were not further purified following deprotection and desalting, but were immediately phosphorylated using polynucleotide kinase (Pharmacia) under the conditions described by Maniatis, T., et al., Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory, publisher, Cold Spring Harbor, New York (1982). Ten nanograms of each phosphorylated oligonucleotide were separately added to 500 ng of single-stranded template. Each mixture was heated to 70°C briefly, cooled to room temperature, and deoxynucleotide triphosphates, 10 units of reverse transcriptase (Life Sciences) and reverse transcriptase buffer were added to make a final reaction volume of 10 ul. After a one-hour incubation at 37°C, ATP, dithiotreitol and bovine serum albumin were added to approximate the recommended reaction conditions for T4 DNA ligase (New England Biolabs) which was added (400 units) for an additional 15-minute incubation at 37°C. One fifth of each ligation reaction was used to transform E. coli resulting in approximately 1,000 bacterial colonies from each oligonucleotide. The colonies from each plate were scraped into LB and 20 glycerol stocks were made, each representing a population of piH3MCD2 mutagenized within a different 33 basepair sequence.
Single-stranded template DNA was prepared in strain BW313/p3, derived from BW313 (Kunkel, T.A., Proc. Natl. Acad. Sci. USA 82:488-492 (1985)) by transformation with the RP1-related plasmid p3 (Seed, B., Nuc. Acid Res. 11:2427-2445 (1983)). BW313/p3 allows selection of plasmids containing a suppressor tRNA gene by growth in media containing ampicillin and tetracycline. The plasmid piH3MCD2 was introduced into BW313/p3, and single-stranded DNA was prepared by infecting the plasmid-carrying strain with wild-type M13 virus as described (Levinson, A., et al., J. Mol. Appl. Genetics 2:507-517 (1984)). The single-stranded DNA produced in BW313/p3 has 20-30 uracil residues per template. This allows efficient selection for the DNA strand made in vitro and thus for incorporation of the oligonucleotide (Kunkel, T.A., supra). Incorporation frequency was approximately 75%.
Specific amino acid changes at positions 46, 47, and 48 were made in a similar fashion, except that oligonucleotides with specified coding potential were used instead of degenerate oligonucleotides.

EXAMPLE II CD2 Mutant Selection

Spheroplasts were prepared from bacteria harboring mutagenized piH3MCD2 and fused to COS cells as described in Applicants' co-pending United States patent application Serial No. 160,416, filed February 25, 1988. Forty-eight hours following fusion, the COS cells expressing CD2 were detached from the dish in PBS/1mM EDTA. The COS cells from six 60-mm dishes were then incubated in PBS, 10% calf serum, 0.02% sodium azide (PBS-FBS) containing a 1/1000 dilution of ascites fluid of the negative selection antibody. All antibody incubations were performed in 1 ml of PBS-FBS for 30 minutes on ice and were followed by centrifugation through a cushion of 2% Ficoll in PBS. The cells were then incubated with 5 ug/ml rabbit anti-mouse Ig antibody (Rockland). Two mls of 50% rabbit complement (Pel-Freeze) in DME (GIBCO) were then added and incubated at 37°C for 30 minutes with agitation to prevent clumping. The complement was removed by dilution to 10 mls with PBS/5mM EDTA and centrifugation of the cells through a 5 ml ficoll cushion. The cells were then incubated with a 1/1000 dilution of the positive selection antibody and added to sheep antimouse immunoglobulin (Cooper Biomedical) coated dishes (Wysocki, L.J., and Sato, V.L. Proc. Natl. Acad. Sci. USA 75:2844-2848 (1978)) prepared as described in Applicants' co-pending United States patent application Serial No. 160,416, filed February 25, 1988. After allowing an hour for the cells to attach, the nonadherent cells were gently washed away, and DNA was prepared from a Hirt supernatant of the adherent cells. The recovered DNA was transformed into E. coli MC1061/p3, and the resultant colonies were either subjected to further rounds of selections, or analyzed directly. For direct analysis, plasmid DNA prepared from individual colonies was used to transfect COS cells by a DEAE dextran procedure as described in Applicants' co-pending United States patent application Serial No. 160,416, filed February 25, 1988. The mutant phenotypes were assayed by sequential indirect immunofluorescence using first the negative selection antibody followed by the positive selection antibody.
When the total extracellular domain of CD2 was to be mutagenized, bacteria from all 20 pools were combined to give a randomly substituted preparation. For directed mutation, the bacteria from only one or two pools were used.
A partial panel of monoclonal antibodies was obtained through the Third International Workshop on Leukocyte Typing.

EXAMPLE III DNA sequence Analysis and Rosetting of CD2 Mutants

The DNA sequence of the mutants was determined on single-stranded templates or alkali denatured plasmid DNA by the dideoxynucleotide method (Sanger, F., et al., Proc. Natl. Acad. Sci. USA 74:5463-5467 (1977)). For mutants obtained using randomly mutated piH3MCD2, the entire extracellular domain of CD2 was sequenced. For mutants obtained by directed mutagenesis a 200 bp portion of the cDNA containing the mutated region was sequenced. In all cases the mutations fell within the span of a single oligonucleotide as expected.
Forty-eight hours after transfection, a 2% suspension of Peterson erythrocytes was allowed to settle onto adherent COS cells for one-half hour at room temperature. Excess erythrocytes were gently removed and the phenotype was scored by microscopic inspection. All rosette phenotypes were assayed on at least three separate occasions with wild-type and negative controls. In some instances, expression of surface CD2 was confirmed by indirect immunofluorescence following exposure to erythrocytes.

Wild-type rosettes are characterized by tight binding of erythrocytes to the COS cells. Partial rosetting has fewer erythrocytes more loosely bound than do wild-type rosettes. Lack of rosetting means that the mutant is indistinguishable from a negative control (i.e., CS8 - Expressing COS cells). Table 2 summarizes the various mutant substitutions giving rise to rosette patterns. Amino acid substitutions separated by commas are present on the same molecule.

TABLE 2

Wild Type Rosette	Partial Rosette	No Rosette
Lys47Ala/Asn	Lys46Ala/Asn	Lys48Ala/Met/Glu
Gln51Arg	GlnPhe51-2ArgSer	Gln51Leu
	Phe52Val,Glu55Gly	Gln51Pro
TyrGln140-1AsnHis	Tyr91Asp	Thr93Ser,Gly95,Val
	Asp92His	Asn97lle

EXAMPLE IV CD4 Mutant Selection

The general method described herein by which surface antigen epitope loss mutants are isolated has been applied to the CD4 antigen with the modification that spheroplasts were prepared from bacteria harboring mutagenized plasmid piH3MCD4. The CD4 antigen is of particular interest because it serves as the T-cell binding site for HIV.
Isolation of epitope - loss mutants was carried out as follows: A CD4 cDNA isolated from the HPB-ALL cell line was mutagenized by annealing degenerate oligonucleotides, prepared as described in Example I, to a deoxyuridine substituted single stranded template. The mutagenized population of plasmids was introduced into COS cells by spheroplast fusion. Two days after fusion, the cells were harvested and sequentially incubated with an anti-CD4 monoclonal antibody, rabbit anti-mouse Ig antisera and rabbit complement. This resulted in selection against the cells bearing determinants recognized by the monoclonal antibody. In order to avoid the isolation of rearrangements which completely eliminate CD4 expression, the cells which remained after complement lysis were incubated with pooled anti-CD4 monoclonals and the cells were allowed to adhere to goat anti-mouse Ig coated petri dishes. Plasmid DNA recovered from cells adhering to the antibody coated plates was used to transform E. coli, amplified, and reintroduced into COS cells. After three rounds of selection in COS cells, plasmid preparations from individual bacterial colonies were transfected into COS cells and assayed two days later by indirect immunofluorescence. Plasmids which encoded CD4 molecules that failed to react with the monoclonal used for complement fixation but which retained other CD4 determinants were then sequenced. Figure 4 shows the primary sequence locations of amino acid whose replacement leads to loss of binding the indicated antibodies. Table 3 indicates the amino acid which is substituted in each case. The amino acid variants are referred to by a wild type residue-position-new residue convention in Table 3 and throughout the text, e.g., a variant carrying a substitution of tyrosine for the serine normally found at position 18 is referred to as Ser18Tyr.
Most of the mutations which were isolated were nucleotide substitutions. An insertion and a deletion were also isolated. These two rearrangements are apparently unrelated to the oligonucleotide mutagenesis but instead result from passage through COS cells. It is known that transfection of COS cells results in a mutation rate of about 1%. The majority of the lesions are insertions and deletions.

Epitope Locations

Most of the mutants, including several selected using antibodies which can potently block gp120-CD4 interaction, encode amino acid substitutions in the amino-terminal, Ig V related, domain of CD4. This domain has several of the cardinal features of an Ig domain, two cysteines, an arginine and a glutamic acid, which in an immunoglobulin form a disulfide bond and a salt bridge respectively and a conserved tryptophan residue, making it extremely likely that the amino-terminus of CD4 folds in the same fashion as an Ig V domain. Based on the hypothesis that this domain of CD4 does fold in a fashion resembling an Ig V domain, a model of the locations of the amino acid substitutions in a folded structure can be made. Modelling allows predictions about gp120 interaction with this domain of CD4 to be made. Figure 3 shows the alignment between CD4 and an Ig V domain. Figure 4 shows a stereo picture of the folding of this same Ig V domain as well as the locations of some landmark residues.
The 66.1 antibody selects mutations causing amino acid variation of the amino terminus of CD4 (Figure 4 and Table 3). The substitutions are separated by up to 23 residues, suggesting that 66.1 recognizes an epitope which is formed by the three dimensional folding pattern of the protein and not by the primary amino acid sequence. The mutations fall into two regions which would correspond to the first and second hypervariable regions of an Ig V segment. These two regions are near each other in a folded immunoglobulin although they are fairly distant in the primary sequence (Figures 2 and 3). This result strongly supports the idea that the amino terminal domain of CD4 folds in a similar fashion to Ig V domains.
The ability of anti-CD4 antibodies to block the binding of other anti-CD4 antibodies also supports the folding pattern depicted in Figure 3. For example the mutants selected using the VIT4 and 13B.8.2 antibodies encode amino acid substitutions separated by about 70 residues from those encoded by G19-2 selected mutants (Figure 1 and Table 1). Both 13B.8.2 and VIT4 can substantially block the binding of G19-2; this would be predicted from the Ig folding pattern.
The 66.1 and G19-2 antibodies are unable to block CD4-gp120 interaction even at many time the saturating concentration. This means that gp120 interacts with CD4 in a fashion that does not involve the face of the molecule which would be formed by the B and C strand homologs of CD4 (Figure 3). Binding of either Leu3a or OKT4A can completely block binding of the other antibody. Neither antibody's binding is able to effectively eliminate binding of the 66.1 and G19-2 antibodies. The mutant selected using OKT4A encodes a substitution for a residue in the D strand of an Ig fold; the Leu3a selected mutants encode variants of the C′ and C˝ strands. The configuration of the B, C C′, C˝ and D strands in an Ig fold leads to the prediction that Leu3a and T4A interacts with the face of the domain formed by the D, E and C˝ strands homologs of CD4. Both T4A and Leu3a block CD4-gp120 interaction at very low antibody concentration. If OKT4A and Leu3a block access of gp120 to a site on the amino terminal domain of CD4, the site must be on the D, E and C˝ strand homologous regions of CD4.
Binding of the VIT4 antibody to CD4 does not interfere with the binding of either Leu3a or T4A. This is not surprising given the VIT4 epitope location (Figure 4), however VIT4 is as potent as Leu3a and T4A at blocking CD4-gp120 dependent syncytium formation. VIT4 might indirectly block access of gp120 to a binding site on the D, E, C˝ strand or might interfere with a spatially distinct binding site, either by blocking access indirectly or by occupying the binding site. The fact that G19-2 competes with VIT4 for binding to CD4 but does not block CD4-gp120 interaction suggests that the ability of VIT4 to block gp120 is indirect. Phylogenetic conservation of HIV susceptibility without conservation of the VIT4 epitope supports the notion that this epitope is not directly involved in HIV binding.
The MT151 antibody blocks interactions with CD4 in a similar fashion to VIT4; it blocks gp120 interaction but not OKT4A of Leu3a binding. The substitutions encoded by MT151 epitope loss mutants are found 5 residues and 77 residues carboxy terminal to the VIT4 epitope loss associated substitution. The MT151 antibody clearly recognizes an epitope formed by the conformation of the folded protein. This epitope overlaps with the VIT4 and 13B.8.2 epitope region but also includes the carboxy terminal part of the second domain. The MT151 epitope places the carboxy terminus of the second disulfide bonded domain of CD4 in close proximity to the carboxy terminus of the amino terminal Ig-like domain and strengthens the possibility that VIT4 and MT151 block CD4-gp120 interaction by blocking access to a site on the second domain, not the first domain.

Effect of Mutations on Syncytium Formation

The mutants allow some of the predictions made from the epitope locations to be tested. One manifestation of CD4-gp120 interaction is the formation of multinucleate, giant cells. These syncytia are at least partly responsible for the cytopathic effects of viral infection. Syncytia can be formed by cells expressing only the HIV env gene product gp160 interacting with cells expressing CD4. We were unable to induce syncytia in transfected COS cells however HeLa cells proved to be very effective at cell-cell fusion. The mutants were each tested for their ability to collaborate in syncytia formation, by transfection of HeLa cells, followed by infection with a vaccinia virus recombinant which expresses HIV gp160. HeLa cell lines which stably express some of the CD4 mutants were created using a retroviral vector which carries G-418 resistance. Single clones or pooled, G-418 resistant cells were expanded and tested for syncytium formation following infection with a vaccinia virus expressing gp160. The same results were obtained with the transient assay, stably expressing clones or pooled G-418 cell lines.
Most of the amino acid variants were as effective as wild-type CD4 in supporting syncytia formation. In contrast, replacements of amino acids 39, 44, 45 and 46 had a striking effect on the ability of CD4 to induce syncytia as did the insertion mutation which was isolated using MT151 selection. Several of the substitutions seemed likely to cause a local denaturation of the protein. For example, the variant selected using the T4/18T3A9 antibody had a proline replacing the glutamine normally found at position 39. Proline has a much more constrained backbone than glutamine and might be expected to alter the local folding structure. Similarly the variants selected using OKT4D, Gly46Arg and Gly46Glu, replace glycine with bulky, charged residues and might denature the C˝ strand homolog of CD4.
Although the substitutions which affect syncytium formation may disrupt protein folding, several observations suggest that the effects on syncytium formation are due to short range changes in the protein structure. Substitutions, similar to those which eliminate CD4 mediated syncytium formation, when at other locations, do not have the same effect; the Gln39Pro variant has only a moderately reduced ability to support syncytium formation, the Gly37Glu and Gln164Pro variants are as effective as wild-type CD4 in participating in syncytium formation. The Lys45Asn, Gly46Val double substitution variant selected using Leu3a is not obviously disruptive but does eliminate syncytium formation.
The effect of the insertion mutation, isolated using MT151, on the structure of CD4 is hard to assess. It results in a large insertion, 13 amino acids, near the end of the second domain. The point mutations selected using MT151 demonstrate that CD4 folds in a fashion such that this portion of the second domain is very close to the carboxy terminus of the first domain. It is possible that the insertion has very indirect effects on CD4 folding and alters the first domain.
Mutations which disrupt syncytia formation could be of at least two classes. It seems certain that mutations which eliminated the ability of CD4 to bind gp120 would eliminate the ability to induce syncytia formation. Another class of mutation is also possible; mutations which eliminate syncytia formation but not binding.

Effect of Mutations on HIV Binding

The effect of the mutations on the ability of CD4 to bind HIV was assessed using an indirect immunofluorescence assay. Concentrated virus particles were incubated with COS cells expressing the different mutants and bound particles were detected using human sera with a high anti-gp120 titer. Binding was quantitated by analysis of cells on a cytofluorometer. In each case, the expression of CD4 on the cell surface was quantitated in parallel using anti-CD4 monoclonals and indirect immunofluorescence. The ability of the variant CD4 molecules to bind virus was consistent with their relative effectiveness in syncytia induction. The mutations which eliminated syncytia formation also eliminated binding of virus particles. Mutations which profoundly reduced the ability of CD4 to support cell fusion likewise reduced their ability to bind gp120. This shows that the primary effect of the mutations is on the ability of CD4 to bind gp120 and not on their ability to induce syncytia.

Discussion

The effects of the various mutations on the ability of CD4 to bind gp120 are largely consistent with the predictions made from the epitope locations. Mutations which are likely to disrupt the C′ strand have altered behavior in binding and syncytia formation assays. Mutations which alter the C˝ strand eliminate the ability of CD4 to bind virus. The mutant selected using the OKT4A antibody might have been expected to behave differently from wild-type CD4 in the syncytia formation assay. This mutant encodes an amino acid replacement in the D strand homologous region of the CD4 cDNA. The substitution encoded by the mutant alters a sequence pattern which is conserved between mouse, rat and human CD4 (mouse-SKKG, rat-SRKN, human-SRRS, OKT4A⁻-SRRR). It seems likely that this amino acid replacement would alter the predicted D strand of CD4, however it does not alter the binding of HIV. In addition the OKT4A epitope is poorly conserved in primate species which are infectable by HIV, suggesting again that the ability of OKT4A to block HIV binding is indirect presumably by blocking access to the Leu3A epitope region.
The VIT4 epitope is also not as well conserved as HIV infectability in primate species and the mutation selected using this antibody does not affect HIV binding. VIT4 probably also blocks HIV binding by indirectly preventing gp120 from interacting with sequences distinct from the antibody recognition site. The mutations selected using this antibody fall in what would correspond to the fourth hypervariable region of an Ig V domain. It is possible that VIT4 blocks access to that portion of the HIV binding site which we have identified as the Leu3a epitope. Such an explanation is not obvious from the proposed relationship between the two epitope regions and from the fact that VIT4 does not interfere with the binding of Leu3a and OKT4A. The MT151 epitope, which overlaps the VIT4 epitope on the first domain, also reaches the carboxy terminus of the second domain. This epitope connects the VIT4 epitope region more closely to the second domain than to the D, E, C˝ face of the first domain. The ability of VIT4 and MT151 to block gp120-CD4 interaction therefore raises the possiblity of a second domain HIV interaction site.
Anti-idiotype antisera which recognize the Leu3a antibody are able to neutralize diverse isolates of HIV. This led to the suggestion that this antibody must recognize an important determinant of the HIV binding site. The concordance between expression of the Leu3a epitope and susceptibility to HIV in primate species supports the importance of the Leu3a epitope to HIV binding. Direct evidence of identity between the recognition site for Leu3a and a site which is important for the binding of HIV is provided. CD4 variants which are altered in recognition by Leu3a are also altered in their ability to bind HIV. The Leu3a epitope resides on the segment of CD4 which would correspond to the C′ and C˝ strands of an Ig V domain. It may be possible to design reagents which mimic the structure of this portion of the CD4 molecule. Short peptides which comprise the C˝ strand and its flanking regions might bind to gp120 with sufficient avidity to prevent gp120 interaction with CD4. Replacement of the C′ and C˝ strands of an authentic Ig V region with the CD4 cognates might similarly result in a structure which could interfere with HIV replication and, or pathology. Such reagents would be important therapeutic agents in the treatment of AIDS.

The primary amino acid sequence of the human CD4 protein is shown in Figure 4. The sequence shown below this represents the mouse sequence where it differs from the human sequence. The underlining indicates the extent of the transmembrane domain of CD4. The bar above the sequence represents the proposed HIV binding site of the CD4 protein. The 16 different antibodies used in the positive selection step are shown along the left-hand margin. The * symbol above the primary sequence indicates that a mutant isolated using a particular antibody has a substitution at that position. Substitutions at amino acid positions 39 and 46 will affect HIV binding. These positions are also indicated in Figure 4. Table 3 presents a summary of the antibodies used in the negative selection step and of the CD4 mutant collection isolated using the methods of this invention.

TABLE 3

Antibody used for negative selection	Amino acid replacement (indicated by normal residue-position-new residue)
Leu3a	K34E*	Lys34Glu
Leu3a	N38K and G46D	Asn38Lys and Gly46Asp
Leu3a	Q39P and T44S	Gln39Pro and Thr44Ser
Leu3a	K45N and G46V	Lys45Asn and Gly46Val
T4	P47S	Pro47Ser
94b1	T44P	Thr44Pro
OKTD	G46D	Gly46Asp
OKT4D	G46R	Gly46Arg
T4/18T3A9	Q39P	Gln39Pro
F101-5	G37E	Gly37Glu
G19-2	S18Y	Ser18Tyr
G19-2	Q19K	Gln19Lys
66.1	Q19H and H26R	Gln19His and His26Arg
66.1	S22G and I23V and H26L	Ser22Gly and Ile23Val and His26Leu
13B.8.2	E76K	Glu76Lys
VIT4	Q78L	Gln78Leu
OKT4A	S59R	Ser59Arg
BL/10T4	G110A	Gly110Ala
BL/10T4	P121H	Pro121His
OKT4E	P121H	Pro121His
OKT4B	Q164P	Gln164Pro
NUTH-1	A216G	Ala216Gly
MT321	V325I	Val325Ile

* Single letter codes for amino acids are used in this column.

Equivalents

Those skilled in the art will recognize, or be able to ascertain, using no more than routine experimentation, many equivalents to the specific embodiment of the invention described herein. Such equivalents are intended to be encompassed by the following claims.

Claims

1. A method of isolating gene mutants, comprising the steps of:

a. transfecting mammalian cells with a vector into which is ligated a structural gene of interest;

b. treating transfected mammalian cells with a first monoclonal antibody directed against an epitope of the protein encoded by the structural gene and with complement, under conditions appropriate for cells expressing the epitope to bind to the first monoclonal antibody and complement fixation and lysis of the cells expressing the epitope to occur;

c. treating the remaining non-lysed mammalian cells with a second monoclonal antibody directed against a second epitope of the structural gene;

d. causing the treated mammalian cells to adhere to a substrate coated with antisera recognizing the second monoclonal antibody; and

e. recovering plasmid DNA from the adherent cells.

2. A method of Claim 1, wherein the mammalian cell is a COS cell.

3. A method of Claim 1, wherein the structural gene of interest is a cell-surface antigen.

4. A method of Claim 3, wherein the cell surface antigen is the CD2 antigen of T-lymphocytes or the CD4 antigen of T-lymphocytes.

5. A method of identifying epitopes of cell surface proteins, comprising the steps of:

b. treating the mammalian cells with a first monoclonal antibody directed against an epitope of the protein encoded by the structural gene and with complement, under conditions appropriate for cells expressing the epitope to bind to the first monoclonal antibody and in complement fixation and lysis of the cells expressing the epitope to occur;

d. causing the treated mammalian cells to adhere to a substrate coated with antisera recognizing the second monoclonal antibody;

e. recovering plasmid DNA from the adherent cells;

f. amplifying recovered plasmid DNA in an appropriate host cell;

g. transfecting mammalian cells with a vector of step (a) in which the structural gene is the amplified recovered plasmid DNA;

h. repeating steps (b) - (g) as appropriate;

i. transfecting mammalian cells with plasmid DNA resulting from the process of steps (a) - (h); and maintaining transfected cells under conditions appropriate for expression of the plasmid DNA to occur; and

j. detecting binding of the product of expression of the plasmid DNA to the first monoclonal antibody.

6. A method of identifying epitopes of cell surface proteins, comprising the steps of:

a. transfecting COS monkey cells with the vector piH3MCD2 into which is ligated a gene encoding at least a portion of the cell surface protein;

b. culturing transfected COS cells for about 48 hours;

c. harvesting transfected cells and treating them sequentially with a negative selection antibody, rabbit anti-mouse immunoglobulin antibody, and complement, under conditions appropriate for cells expressing the cell surface protein and the negative selection antibody, complement fixation and lysis of said cells to occur;

d. harvesting the unlysed cells and treating the unlysed cells with a positive selection antibody;

e. causing the unlysed cells to adhere to a substrate coated with sheep anti-murine immunoglobulin;

f. removing nonadherent cells;

g. preparing and recovering a DNA extract of the adherent cells;

h. transforming the recovered DNA extract into E. coli MC1001/p3; and

i. subjecting the resultant bacterial colonies to further rounds of selections, as in steps (a) - (h) or to direct analysis of ability to bind to the negative selection antibody.

7. A method of Claim 6 wherein the structural gene is a cell-surface antigen of T-lymphocytes and B-cells.

8. A method of Claim 6, wherein the negative selection antibody and the positive selection antibody are selected from the group consisting of the antibodies in Table 1.

9. In a method of isolating epitopic variants of cell surface antigens using materials coated with antibodies directed against the cell surface antigen, the improvement comprising:

a. treating the cells expressing the antigen in such a manner that cells expressing a selected epitope are lysed; and

b. treating the remaining non-lysed cells with a monoclonal antibody recognizing a second, distinct epitope of the cell surface antigen.